This repository contains Eigen-based C++ implementation of a Multi-Layer Perceptron (MLP) and a Gated
Recurrent Unit (GRU) network. In addition, the repository also provides example scripts to export PyTorch
network parameters to .txt files. These parameters can then be loaded into the C++ networks.
This README assumes that $PROJECT_DIR refers to the root of this repository.
To build the shared library:
cd $PROJECT_DIR
mkdir build && cd build
cmake .. -DCMAKE_BUILD_TYPE=Release
make
This will generate a shared library at $PROJECT_DIR/lib. The master branch
of the repository only contains the header and source files necessary for building the
library.
For example usage, switch to example branch first.
cd $PROJECT_DIR
git checkout example
After rebuilding the library and executables, as shown above,
you can execute the examples provided in
$PROJECT_DIR/examples/<robot>_actuation.cpp like so:
cd $PROJECT_DIR/build
./<robot>_actuation
Here, <robot> can either be a1 or boxy.
Example code to export PyTorch network parameters is provided in
$PROJECT_DIR/scripts/torch_export.py. To
execute this, creating a virtual environment is recommended:
cd $PROJECT_DIR
python3 -m venv venv
source venv/bin/activate
pip install -e .
After this, execute the script using:
python scripts/torch_export.py
This will create the $PROJECT_DIR/scripts/exported_paramters directory, and the network
parameters will be stored here.
As of yet, you can only load network parameters and perform a forward pass with the MLP and GRU
implementations. The MLP also allows computation of network Jacobian using the gradient()
function.
The current implementation builds upon dynamic Eigen matrices. A static implementation could possibly offer better performance. This is currently not a priority since most of the forward passes for smaller networks can be performed in microseconds. However, any contribution from the community in this direction is highly encouraged.
- Siddhant Gangapurwala siddhant@gangapurwala.com